9-geometry of lp

Upload: rafal-dembowski

Post on 14-Apr-2018

225 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/27/2019 9-Geometry of LP

    1/22

    DSC1007BusinessAnalytics

    TheGeometryofLinearProgramming

    Na#onalUniversityofSingapore

    Lecture8

  • 7/27/2019 9-Geometry of LP

    2/22

    OriginofOptimizationModeling Optimization:What?When?Why? Optimizationprocess:

    OptimizationModel:3majorcomponents Objective Decisionvariables Constraints

    Standardformulationofoptimizationmodels Examples:

    production,transportation,investment

    2

    PracticalProblem

    OptimizationModel

    Solution

    Recap:LinearOptimization

  • 7/27/2019 9-Geometry of LP

    3/22

    Howtovisualizealinearprogramgeometrically? Decisionvariables? Constraints? Objective?

    HowtoNindtheoptimalsolutiongeometrically? Anexample:

    3

    GeometryofLinearProgramming

  • 7/27/2019 9-Geometry of LP

    4/22

    Eachdimensionrepresentsadecisionvariableann-dimensionalspace

    Eachpointinthespacerepresentsaparticularsolution

    4

    X1

    X2

    0

    GeometryofLPDecisionVariables

    (2, 3)3

    2

  • 7/27/2019 9-Geometry of LP

    5/22

    Constraint:x10 Inequalityconstraintx10deNinesahalfspace

    5

    X1

    X2

    X1 0

    GeometryofLPConstraints

    0

    x1 = 0

  • 7/27/2019 9-Geometry of LP

    6/22

    Constraint:x20 Inequalityconstraintx20deNinesahalfspace

    6

    X1

    X2

    0

    X2 0

    GeometryofLPConstraints

    x2 = 0

  • 7/27/2019 9-Geometry of LP

    7/22

    Constraint:2x1+x23 Inequalityconstraint2x1+x23deNinesahalfspace

    7

    X1

    X2

    0

    2x1 + x23

    GeometryofLPConstraints

    2x2 +x2 = 3(0, 3)

    (1.5, 0)

  • 7/27/2019 9-Geometry of LP

    8/22

    Constraint:x1+2x23 Inequalityconstraintx1+2x23deNinesahalfspace

    8

    X1

    X2

    0

    x1 + 2x23

    GeometryofLPConstraints

    (3, 0)

    (0, 1.5)

    x2 +2x2 = 3

  • 7/27/2019 9-Geometry of LP

    9/22

    TheintersectionofalltheconstraintsFeasibleRegion

    9

    X1

    X2

    0

    GeometryofLPConstraints

  • 7/27/2019 9-Geometry of LP

    10/22

    FeasibleRegion:Thesetofalltheallowedsolutions;aregion(polygon)boundedbytheconstraints EachequalityconstraintdeNinesaline EachinequalityconstraintdeNinesahalf-space

    ExtremePoints:Cornerpointsontheboundaryofthefeasibleregion.E.g.,(0,0),(1.5,0),(0,1.5),and(1,1)

    Infeasibleproblem:Aproblemwithanemptyfeasibleregion Redundantconstraint:Addingorremovingtheconstraintdoes

    notaffectthefeasibleregion

    10

    X1

    X2

    0

    GeometryofLPConstraints

  • 7/27/2019 9-Geometry of LP

    11/22

    Objective:maximizex1+x2 Isoquant:Alineonwhichallpointshavethesameobjective

    value;allpointsareequallygoodontheobjectivefunction.

    11

    X1

    X2

    0

    x1

    + x2

    = 0

    GeometryofLPObjective

  • 7/27/2019 9-Geometry of LP

    12/22

    Isoquant:x1+x2=1

    12

    X1

    X2

    0

    x1

    + x2

    = 1

    GeometryofLPObjective

    (1, 0)

    (0, 1)

  • 7/27/2019 9-Geometry of LP

    13/22

    Isoquant:x1+x2=1.5

    13

    X1

    X2

    0

    x1

    + x2

    = 1.5

    GeometryofLPObjective

  • 7/27/2019 9-Geometry of LP

    14/22

    Isoquant:x1+x2=3

    14

    X1

    X2

    0

    x1

    + x2

    = 3

    GeometryofLPObjective

  • 7/27/2019 9-Geometry of LP

    15/22

    OptimalSolution:Thebestfeasiblesolution Theorem:ForanyfeasibleLPwithaNiniteoptimalsolution,

    thereexistsanoptimalsolutionthatisanextremepoint

    15

    X1

    X2

    0

    Maxx1

    + x2

    OptimalSolution:

    (1,1)

    GeometryofLPOptimalSolution

    x1 + x2= 2

  • 7/27/2019 9-Geometry of LP

    16/22

    OptimalsolutionsmayOTbeunique

    16

    X1

    X2

    0

    Max 2x1

    + x2

    GeometryofLPOptimalSolution

  • 7/27/2019 9-Geometry of LP

    17/22

    OptimalsolutionmayOTbeNinite

    17

    X1

    X2

    0

    Maxx1

    + x2

    X1 - 2X2 2

    -X1 + 2X2 2

    GeometryofLPOptimalSolution

    (2, 0)

    (0, 1)

  • 7/27/2019 9-Geometry of LP

    18/22

    Binding(oractive)constraints:TheconstraintsthataresatisNiedatequalityattheoptimalsolution

    AllequalityconstraintsarebindingbydeNinition Non-binding(orinactive)constraintsaresatisNiedat

    strictinequalityattheoptimalsolution

    Theinequalitylevel(=RHSLHS)isknownastheslack BindingconstraintshavezeroslackbydeNinition

    18

    GeometryofLPOptimalSolution

  • 7/27/2019 9-Geometry of LP

    19/22

    Activeconstraintsaretheonesthatpassthroughtheoptimalsolution. Inactiveconstraintsaretheonesthatdonotpassthroughtheoptimal

    solution.

    Whyarebindingconstraintsimportant?

    19

    Constraints Binding? Slack

    X1 + 2X2 3 Yes 0

    2X1 + X2 3 Yes 0

    X1 0 No 1

    X2 0 No 1

    GeometryofLPOptimalSolution

    Optimal

    (1, 1)

  • 7/27/2019 9-Geometry of LP

    20/22

    Decisionvariables

    Eachdecisionvariableformsadimension

    ndecisionvariablesdeNineann-dimensionalspace Asolutionisapointinthespace

    Constraints EachequalityconstraintdeNinesahyper-plane(lineina2-Dspace) EachinequalityconstraintdeNinesahalf-space AllconstraintscollectivelydeNinethefeasibleregion

    Objective TheobjectivefunctiondeNinesisoquantsandadirectioninthespace

    ToNindingtheoptimalsolution,pushalongthedirectiondeNinedbytheobjectiveuntilwereachtheboundaryofthefeasibleregion

    Attheoptimalsolution,someconstraintsarebinding(oractive)whileothersarenot

    20

    GeometryofLPSummary

  • 7/27/2019 9-Geometry of LP

    21/22

    AinNinitenumberoffeasiblesolutions

    ANinitenumberofextremepoints Howmany? (+ )

    Anextremepointistheoptimalsolution

    TheSimplexalgorithmdevelopedbyDantzig

    21

    GeometryofLPBeneNits

  • 7/27/2019 9-Geometry of LP

    22/22